AI Architect & Tech Lead for Mahjong Game AI

About the role

We are looking for a hands‑on AI Architect and Tech Lead to build an intelligent Mahjong companion for an online social gaming platform. The AI will play at a strong level, explain its moves, and operate within a strict 2‑second response budget.

Key responsibilities

• Design and implement the core Mahjong‑playing algorithm using reinforcement learning, imitation learning, search (MCTS) or a hybrid approach.
• Integrate a reasoning layer powered by a large language model to generate move explanations.
• Own the end‑to‑end technical architecture, including the game model, valid‑action mask, win detection and API contract with the game engine.
• Optimize inference latency through profiling, batching and caching to meet the 2‑second budget.
• Define data requirements, build training and calibration pipelines from hand‑history logs, and create an evaluation harness for strength and explanation quality.
• Set up LLM observability with Langfuse and lead a small pod of an AI Engineer and client stakeholders in a Scrum process.
• Communicate technical decisions and trade‑offs to the client’s CTO and product owners.

Required profile

• Strong background in game AI and sequential decision‑making for imperfect‑information games.
• Hands‑on experience building RL, imitation‑learning or search‑based agents.
• Proficiency in Python and production‑grade software engineering (APIs, testing, CI).
• Experience deploying low‑latency ML models on AWS.
• Ability to lead a small technical team and interact with senior client engineers.

Required skills

• Python
• Reinforcement learning (RL)
• Imitation learning
• Monte‑Carlo Tree Search (MCTS) / self‑play
• Large language model integration
• API development
• Testing and CI pipelines
• AWS for ML workloads
• Low‑latency inference optimization (profiling, batching, caching)
• Langfuse or similar LLM observability tools